Seattle City Energy Consumption - Part 1, EDA

  • 1. Introduction
    • 1.1 Columns Summary
    • 1.2 Importing Libraries
    • 1.3 Loading Data
    • 1.4 Deleting Repeated and Unused Columns
    • 1.5 Selecting Building Type
    • 1.6 Missing Values for ENERGYSTARScore
  • 2. EDA
    • 2.1 PrimaryPropertyType and LargestPropertyUseType
    • 2.2 Building Type
    • 2.3 Compliance Status, Default Data, Outliers
    • 2.4 Seattle Neighborhood
    • 2.5 Seattle Maps
    • 2.6 Number of Buildings
    • 2.7 "Weather Normalized" Columns
  • 3. Target Data and EnergyStar Score
    • 3.1 Target Columns Distribution
    • 3.2 Energy Ratios
    • 3.3 Buildings Surface and Geometry
    • 3.4 Correlation Bewteen the Variables
    • 3.5 Construction Decade
    • 3.6 ENERGYSTAR Score
  • 4.Final Columns Selection

1. Introduction

1.1 Columns Summary

  • DataYear, State, et City only contains a single value each
  • "comments" does not contain any value
  • The target columns either be SiteEnergyUse and TotalGHGEmissions or theire per unit of surface counterparts, SiteEUI and GHGEmissionsIntensity
  • WN stands for weather normalized, the average value over 30 years

1.2 Importing Libraries

In [1]:
import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
import plotly.offline as pyo

%matplotlib inline
plt.style.use('seaborn')
pyo.init_notebook_mode()

1.3 Loading Data

In [2]:
df = pd.read_csv('2016_Building_Energy_Benchmarking.csv')
df.head()
Out[2]:
OSEBuildingID DataYear BuildingType PrimaryPropertyType PropertyName Address City State ZipCode TaxParcelIdentificationNumber ... Electricity(kWh) Electricity(kBtu) NaturalGas(therms) NaturalGas(kBtu) DefaultData Comments ComplianceStatus Outlier TotalGHGEmissions GHGEmissionsIntensity
0 1 2016 NonResidential Hotel Mayflower park hotel 405 Olive way Seattle WA 98101.0 0659000030 ... 1.156514e+06 3946027.0 12764.52930 1276453.0 False NaN Compliant NaN 249.98 2.83
1 2 2016 NonResidential Hotel Paramount Hotel 724 Pine street Seattle WA 98101.0 0659000220 ... 9.504252e+05 3242851.0 51450.81641 5145082.0 False NaN Compliant NaN 295.86 2.86
2 3 2016 NonResidential Hotel 5673-The Westin Seattle 1900 5th Avenue Seattle WA 98101.0 0659000475 ... 1.451544e+07 49526664.0 14938.00000 1493800.0 False NaN Compliant NaN 2089.28 2.19
3 5 2016 NonResidential Hotel HOTEL MAX 620 STEWART ST Seattle WA 98101.0 0659000640 ... 8.115253e+05 2768924.0 18112.13086 1811213.0 False NaN Compliant NaN 286.43 4.67
4 8 2016 NonResidential Hotel WARWICK SEATTLE HOTEL (ID8) 401 LENORA ST Seattle WA 98121.0 0659000970 ... 1.573449e+06 5368607.0 88039.98438 8803998.0 False NaN Compliant NaN 505.01 2.88

5 rows × 46 columns

In [3]:
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3376 entries, 0 to 3375
Data columns (total 46 columns):
 #   Column                           Non-Null Count  Dtype  
---  ------                           --------------  -----  
 0   OSEBuildingID                    3376 non-null   int64  
 1   DataYear                         3376 non-null   int64  
 2   BuildingType                     3376 non-null   object 
 3   PrimaryPropertyType              3376 non-null   object 
 4   PropertyName                     3376 non-null   object 
 5   Address                          3376 non-null   object 
 6   City                             3376 non-null   object 
 7   State                            3376 non-null   object 
 8   ZipCode                          3360 non-null   float64
 9   TaxParcelIdentificationNumber    3376 non-null   object 
 10  CouncilDistrictCode              3376 non-null   int64  
 11  Neighborhood                     3376 non-null   object 
 12  Latitude                         3376 non-null   float64
 13  Longitude                        3376 non-null   float64
 14  YearBuilt                        3376 non-null   int64  
 15  NumberofBuildings                3368 non-null   float64
 16  NumberofFloors                   3376 non-null   int64  
 17  PropertyGFATotal                 3376 non-null   int64  
 18  PropertyGFAParking               3376 non-null   int64  
 19  PropertyGFABuilding(s)           3376 non-null   int64  
 20  ListOfAllPropertyUseTypes        3367 non-null   object 
 21  LargestPropertyUseType           3356 non-null   object 
 22  LargestPropertyUseTypeGFA        3356 non-null   float64
 23  SecondLargestPropertyUseType     1679 non-null   object 
 24  SecondLargestPropertyUseTypeGFA  1679 non-null   float64
 25  ThirdLargestPropertyUseType      596 non-null    object 
 26  ThirdLargestPropertyUseTypeGFA   596 non-null    float64
 27  YearsENERGYSTARCertified         119 non-null    object 
 28  ENERGYSTARScore                  2533 non-null   float64
 29  SiteEUI(kBtu/sf)                 3369 non-null   float64
 30  SiteEUIWN(kBtu/sf)               3370 non-null   float64
 31  SourceEUI(kBtu/sf)               3367 non-null   float64
 32  SourceEUIWN(kBtu/sf)             3367 non-null   float64
 33  SiteEnergyUse(kBtu)              3371 non-null   float64
 34  SiteEnergyUseWN(kBtu)            3370 non-null   float64
 35  SteamUse(kBtu)                   3367 non-null   float64
 36  Electricity(kWh)                 3367 non-null   float64
 37  Electricity(kBtu)                3367 non-null   float64
 38  NaturalGas(therms)               3367 non-null   float64
 39  NaturalGas(kBtu)                 3367 non-null   float64
 40  DefaultData                      3376 non-null   bool   
 41  Comments                         0 non-null      float64
 42  ComplianceStatus                 3376 non-null   object 
 43  Outlier                          32 non-null     object 
 44  TotalGHGEmissions                3367 non-null   float64
 45  GHGEmissionsIntensity            3367 non-null   float64
dtypes: bool(1), float64(22), int64(8), object(15)
memory usage: 1.2+ MB

1.4 Deleting Repeated and Unused Columns

In [4]:
df.drop(['DataYear', 'City', 'State', 'ZipCode',
        'TaxParcelIdentificationNumber', 'CouncilDistrictCode',
         'Address', 'YearsENERGYSTARCertified', 'Comments',
        'Electricity(kWh)', 'NaturalGas(therms)'],
       axis=1, inplace=True)
In [5]:
df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3376 entries, 0 to 3375
Data columns (total 35 columns):
 #   Column                           Non-Null Count  Dtype  
---  ------                           --------------  -----  
 0   OSEBuildingID                    3376 non-null   int64  
 1   BuildingType                     3376 non-null   object 
 2   PrimaryPropertyType              3376 non-null   object 
 3   PropertyName                     3376 non-null   object 
 4   Neighborhood                     3376 non-null   object 
 5   Latitude                         3376 non-null   float64
 6   Longitude                        3376 non-null   float64
 7   YearBuilt                        3376 non-null   int64  
 8   NumberofBuildings                3368 non-null   float64
 9   NumberofFloors                   3376 non-null   int64  
 10  PropertyGFATotal                 3376 non-null   int64  
 11  PropertyGFAParking               3376 non-null   int64  
 12  PropertyGFABuilding(s)           3376 non-null   int64  
 13  ListOfAllPropertyUseTypes        3367 non-null   object 
 14  LargestPropertyUseType           3356 non-null   object 
 15  LargestPropertyUseTypeGFA        3356 non-null   float64
 16  SecondLargestPropertyUseType     1679 non-null   object 
 17  SecondLargestPropertyUseTypeGFA  1679 non-null   float64
 18  ThirdLargestPropertyUseType      596 non-null    object 
 19  ThirdLargestPropertyUseTypeGFA   596 non-null    float64
 20  ENERGYSTARScore                  2533 non-null   float64
 21  SiteEUI(kBtu/sf)                 3369 non-null   float64
 22  SiteEUIWN(kBtu/sf)               3370 non-null   float64
 23  SourceEUI(kBtu/sf)               3367 non-null   float64
 24  SourceEUIWN(kBtu/sf)             3367 non-null   float64
 25  SiteEnergyUse(kBtu)              3371 non-null   float64
 26  SiteEnergyUseWN(kBtu)            3370 non-null   float64
 27  SteamUse(kBtu)                   3367 non-null   float64
 28  Electricity(kBtu)                3367 non-null   float64
 29  NaturalGas(kBtu)                 3367 non-null   float64
 30  DefaultData                      3376 non-null   bool   
 31  ComplianceStatus                 3376 non-null   object 
 32  Outlier                          32 non-null     object 
 33  TotalGHGEmissions                3367 non-null   float64
 34  GHGEmissionsIntensity            3367 non-null   float64
dtypes: bool(1), float64(18), int64(6), object(10)
memory usage: 900.2+ KB

1.5 Selecting Building Type

Before selecting the building, we are going to simplyfy the building type

COS means City Of Seattle and WA if for the state of Washington

In [6]:
building_dic = {'NonResidential': 'NonResidential',
                'Nonresidential COS': 'NonResidential',
                'Multifamily MR (5-9)': 'Residential',
                'SPS-District K-12': 'School',
                'Campus': 'Campus',
                'Multifamily LR (1-4)': 'Residential',
                'Multifamily HR (10+)': 'Residential',
                'Nonresidential WA': 'NonResidential'}

df['BuildingType'] = df['BuildingType'].map(building_dic)
In [7]:
df = df[df['BuildingType'] != 'Residential'].copy()

1.6 Missing Values for ENERGYSTAR Score

We need to determine if the ENERGYSTAR score is useful. Therefore we are only remove the building where the value is missing

In [8]:
df = df[df['ENERGYSTARScore'].notna()].copy()

2. Exploration Data Analysis

2.1 Primary Property Type and Largest Property Use Type

In [9]:
plt.figure(figsize=(12,12), facecolor="#eaeaf2")
sns.histplot(data=df, y='PrimaryPropertyType', hue='LargestPropertyUseType',
            multiple='stack');
In [10]:
plt.figure(figsize=(12,12), facecolor="#eaeaf2")
sns.boxplot(data=df, y='PrimaryPropertyType', x='SiteEUI(kBtu/sf)');

The building with "Parking" as largestpropertyuse will be change based on info find with GoogleMaps.

In addition, one hospital will be labelled as Senior Care Community and one Senior Care Community will be labelled as Hospital

In [11]:
df.loc[308, 'LargestPropertyUseType'] = "Office"
df.loc[309, 'LargestPropertyUseType'] = "Data Center"
df.loc[2800, 'LargestPropertyUseType'] = "Supermarket/Grocery Store"

df.loc[2095, 'LargestPropertyUseType'] = "Retail Store" #Automibili Italine
df.loc[246, 'LargestPropertyUseType'] = "Office" #metropolitan park
df.loc[488, 'LargestPropertyUseType'] = "Office" # West lake SEDO
df.loc[1306, 'LargestPropertyUseType'] = "Retail Store"# Elliot court

df.loc[1147, 'LargestPropertyUseType'] = "Hotel" #Palladian Hotel
df.loc[2414, 'LargestPropertyUseType'] = "Office" #Talon Northlake

df.loc[601, 'LargestPropertyUseType'] = "Supermarket/Grocery Store"

df.loc[1109, 'LargestPropertyUseType'] = 'Hospital (General Medical & Surgical)' #Ballard Center
df.loc[2482, "LargestPropertyUseType"] = "Senior Care Community"
In [12]:
plt.figure(figsize=(12,18), facecolor="#eaeaf2")
sns.boxplot(data=df, y='LargestPropertyUseType', x='SiteEUI(kBtu/sf)');

We can see on the boxplot that the LargestPropertyUseType will be an important column for the predictions

In addition to the change we've already made, we are going to group a few columns together

In [13]:
df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Distribution Center', 'Warehouse')
df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Refrigerated Warehouse', 'Warehouse')
df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Non-Refrigerated Warehouse', 'Warehouse')

df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Wholesale Club/Supercenter', 'Retail Store')
df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Other', 'Retail Store')

df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Bank Branch', 'Bank/Finance/Courthouse')
df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Financial Office', 'Bank/Finance/Courthouse')
df['LargestPropertyUseType'] = df['LargestPropertyUseType'].replace('Courthouse', 'Bank/Finance/Courthouse')
In [14]:
df.LargestPropertyUseType.value_counts()
Out[14]:
Office                                   416
Warehouse                                191
K-12 School                              125
Retail Store                              83
Hotel                                     73
Worship Facility                          65
Supermarket/Grocery Store                 40
Medical Office                            37
Residence Hall/Dormitory                  19
Senior Care Community                     18
Hospital (General Medical & Surgical)     10
Bank/Finance/Courthouse                    9
Data Center                                4
Multifamily Housing                        4
Name: LargestPropertyUseType, dtype: int64
In [15]:
df = df[df['LargestPropertyUseType'] != 'Multifamily Housing']
In [16]:
drop_list = ['ListOfAllPropertyUseTypes',
             'PrimaryPropertyType',
             'LargestPropertyUseTypeGFA',
             'SecondLargestPropertyUseType',
             'SecondLargestPropertyUseTypeGFA',
             'ThirdLargestPropertyUseType',
             'ThirdLargestPropertyUseTypeGFA']
df.drop(drop_list, axis=1, inplace=True)

2.3 Compliance Status, Default Data, Outliers

In [17]:
plt.figure(figsize=(12,6), facecolor="#eaeaf2")
sns.boxplot(data=df, x='ComplianceStatus', y='SourceEUI(kBtu/sf)')
plt.ylim(0,500);
In [18]:
df.ComplianceStatus.value_counts()
Out[18]:
Compliant                       993
Error - Correct Default Data     88
Non-Compliant                     9
Name: ComplianceStatus, dtype: int64
In [19]:
df = df[df['ComplianceStatus'] == 'Compliant'].copy()
In [20]:
df.Outlier.value_counts()
Out[20]:
Series([], Name: Outlier, dtype: int64)
In [21]:
df.DefaultData.value_counts()
Out[21]:
False    993
Name: DefaultData, dtype: int64

The "Non-compliant" building have a lower energy consumption. However we do not have enough info about the data to use them. We will only keep the data labelled as "Compliant"

Keeping only the "Compliant" building also suppressed the data labelled as outliers and the data where DefaultData were not False

In [22]:
df.drop(['ComplianceStatus', 'DefaultData', 'Outlier'],
       axis=1, inplace=True)

2.4 Seattle Neighborhood

We need to take care of the Neighborhood columns becauses the name are not standardized

In [23]:
df.Neighborhood.value_counts()
Out[23]:
DOWNTOWN                 243
GREATER DUWAMISH         200
LAKE UNION               101
MAGNOLIA / QUEEN ANNE     87
NORTHEAST                 80
EAST                      65
NORTHWEST                 51
BALLARD                   33
NORTH                     31
CENTRAL                   25
DELRIDGE                  24
SOUTHEAST                 21
SOUTHWEST                 18
Northwest                  4
Delridge                   3
Ballard                    3
North                      3
Central                    1
Name: Neighborhood, dtype: int64
In [24]:
df['Neighborhood'] = df['Neighborhood'].str.title()
df['Neighborhood'] = df['Neighborhood'].replace('Delridge.*', 'Delrigde', regex=True)
# all the name starting by "Delridge" are replaced by "Delridge", not necessary after the data selection

2.5 Maps of Seattle

In [25]:
def plot_map(column):
    
    fig = px.scatter_mapbox(df,
                           lat='Latitude',
                           lon='Longitude',
                           color=column,
                           hover_name='PropertyName')
    
    fig.update_layout(mapbox_style='open-street-map')
    
    title = 'map-' + column
    pyo.iplot(fig, title)
In [26]:
plot_map('Neighborhood')
In [27]:
plot_map('SiteEUI(kBtu/sf)')
In [28]:
plot_map('GHGEmissionsIntensity')
In [29]:
df.drop(['Latitude', 'Longitude'],
       axis=1, inplace=True)

2.6 Nombre de batiments

In [30]:
plt.figure(figsize=(12,4), facecolor="#eaeaf2")
sns.countplot(data=df, x='NumberofBuildings', color='#7490c0');

la plupart des terrains n'ont qu'un batiment. Nous allons completer les valeurs manquantes et les zeros par cette valeur

In [31]:
df['NumberofBuildings'].fillna(value=1, inplace=True)
df['NumberofBuildings'].replace(0,1, inplace=True)
In [32]:
df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 993 entries, 0 to 3347
Data columns (total 23 columns):
 #   Column                  Non-Null Count  Dtype  
---  ------                  --------------  -----  
 0   OSEBuildingID           993 non-null    int64  
 1   BuildingType            993 non-null    object 
 2   PropertyName            993 non-null    object 
 3   Neighborhood            993 non-null    object 
 4   YearBuilt               993 non-null    int64  
 5   NumberofBuildings       993 non-null    float64
 6   NumberofFloors          993 non-null    int64  
 7   PropertyGFATotal        993 non-null    int64  
 8   PropertyGFAParking      993 non-null    int64  
 9   PropertyGFABuilding(s)  993 non-null    int64  
 10  LargestPropertyUseType  993 non-null    object 
 11  ENERGYSTARScore         993 non-null    float64
 12  SiteEUI(kBtu/sf)        993 non-null    float64
 13  SiteEUIWN(kBtu/sf)      992 non-null    float64
 14  SourceEUI(kBtu/sf)      993 non-null    float64
 15  SourceEUIWN(kBtu/sf)    993 non-null    float64
 16  SiteEnergyUse(kBtu)     993 non-null    float64
 17  SiteEnergyUseWN(kBtu)   992 non-null    float64
 18  SteamUse(kBtu)          993 non-null    float64
 19  Electricity(kBtu)       993 non-null    float64
 20  NaturalGas(kBtu)        993 non-null    float64
 21  TotalGHGEmissions       993 non-null    float64
 22  GHGEmissionsIntensity   993 non-null    float64
dtypes: float64(13), int64(6), object(4)
memory usage: 186.2+ KB

2.7 "Weather Normalized" Columns

In [33]:
fig, ax = plt.subplots(1,2,figsize=(12,4), facecolor="#eaeaf2")

sns.scatterplot(data=df, ax=ax[0], x='SiteEUI(kBtu/sf)', y='SiteEUIWN(kBtu/sf)')
sns.scatterplot(data=df, ax=ax[1], x='SourceEUI(kBtu/sf)', y='SourceEUIWN(kBtu/sf)');
In [34]:
df.drop(['SiteEUIWN(kBtu/sf)', 'SourceEUIWN(kBtu/sf)', 'SiteEnergyUseWN(kBtu)'],
       axis=1, inplace=True)
In [35]:
df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 993 entries, 0 to 3347
Data columns (total 20 columns):
 #   Column                  Non-Null Count  Dtype  
---  ------                  --------------  -----  
 0   OSEBuildingID           993 non-null    int64  
 1   BuildingType            993 non-null    object 
 2   PropertyName            993 non-null    object 
 3   Neighborhood            993 non-null    object 
 4   YearBuilt               993 non-null    int64  
 5   NumberofBuildings       993 non-null    float64
 6   NumberofFloors          993 non-null    int64  
 7   PropertyGFATotal        993 non-null    int64  
 8   PropertyGFAParking      993 non-null    int64  
 9   PropertyGFABuilding(s)  993 non-null    int64  
 10  LargestPropertyUseType  993 non-null    object 
 11  ENERGYSTARScore         993 non-null    float64
 12  SiteEUI(kBtu/sf)        993 non-null    float64
 13  SourceEUI(kBtu/sf)      993 non-null    float64
 14  SiteEnergyUse(kBtu)     993 non-null    float64
 15  SteamUse(kBtu)          993 non-null    float64
 16  Electricity(kBtu)       993 non-null    float64
 17  NaturalGas(kBtu)        993 non-null    float64
 18  TotalGHGEmissions       993 non-null    float64
 19  GHGEmissionsIntensity   993 non-null    float64
dtypes: float64(10), int64(6), object(4)
memory usage: 162.9+ KB

3. Target Data and EnergyStar Score

3.1 Target Columns Distribution

In [36]:
fig, ax = plt.subplots(1,2,figsize=(12,6))
sns.histplot(data=df, x='SiteEUI(kBtu/sf)', ax=ax[0])
sns.histplot(data=df, x='GHGEmissionsIntensity', ax=ax[1]);
In [37]:
df[df['GHGEmissionsIntensity'] > 15]
Out[37]:
OSEBuildingID BuildingType PropertyName Neighborhood YearBuilt NumberofBuildings NumberofFloors PropertyGFATotal PropertyGFAParking PropertyGFABuilding(s) LargestPropertyUseType ENERGYSTARScore SiteEUI(kBtu/sf) SourceEUI(kBtu/sf) SiteEnergyUse(kBtu) SteamUse(kBtu) Electricity(kBtu) NaturalGas(kBtu) TotalGHGEmissions GHGEmissionsIntensity
35 43 Campus Plant 2 Site Greater Duwamish 1992 1.0 3 494835 0 494835 Office 8.0 221.699997 388.200012 448385312.0 0.0 150476283.0 297909000.0 16870.98 34.09
3264 49940 NonResidential Virginia Mason Medical Center - 2149 East 1920 1.0 8 374466 0 374466 Hospital (General Medical & Surgical) 74.0 162.300003 367.899994 163945984.0 73885472.0 90060497.0 0.0 6330.91 16.91
In [38]:
df[df['SiteEnergyUse(kBtu)'] <= 0]
Out[38]:
OSEBuildingID BuildingType PropertyName Neighborhood YearBuilt NumberofBuildings NumberofFloors PropertyGFATotal PropertyGFAParking PropertyGFABuilding(s) LargestPropertyUseType ENERGYSTARScore SiteEUI(kBtu/sf) SourceEUI(kBtu/sf) SiteEnergyUse(kBtu) SteamUse(kBtu) Electricity(kBtu) NaturalGas(kBtu) TotalGHGEmissions GHGEmissionsIntensity
In [39]:
df[df['TotalGHGEmissions'] <= 0]
Out[39]:
OSEBuildingID BuildingType PropertyName Neighborhood YearBuilt NumberofBuildings NumberofFloors PropertyGFATotal PropertyGFAParking PropertyGFABuilding(s) LargestPropertyUseType ENERGYSTARScore SiteEUI(kBtu/sf) SourceEUI(kBtu/sf) SiteEnergyUse(kBtu) SteamUse(kBtu) Electricity(kBtu) NaturalGas(kBtu) TotalGHGEmissions GHGEmissionsIntensity
513 700 NonResidential IUC- Whole Foods Interbay Magnolia / Queen Anne 2008 1.0 1 57176 0 57176 Supermarket/Grocery Store 31.0 208.800003 0.0 1.252517e+07 0.0 0.0 0.0 0.0 0.00
3206 49784 NonResidential Bullitt Center Central 2013 1.0 6 52000 0 52000 Office 100.0 7.100000 2.0 3.427261e+05 0.0 -115417.0 0.0 -0.8 -0.02

The bullitt center is a low carbon emissions building. It was designed to reduce its carbon footprint. The solar panel on its top allows it to produce more electricity than it consumes.

We are going to remove these outliers

In [40]:
df = df[df['TotalGHGEmissions'] > 0].copy()

3.2 Energy Ratios

To predict the Greenhouse Gas emissions, it is useful to know the source of energy of the buildings

  • The first group of columns will be used for predictions
  • And the second group will only be used in this notebook to understand the data
In [41]:
df['SteamRatio'] = df['SteamUse(kBtu)'] / df['SiteEnergyUse(kBtu)']
df['ElectricityRatio'] = df['Electricity(kBtu)'] / df['SiteEnergyUse(kBtu)']
df['NaturalGasRatio'] = df['NaturalGas(kBtu)'] / df['SiteEnergyUse(kBtu)']
In [42]:
df['SteamRatio10'] = (df['SteamRatio'] * 100 + 9)  // 10
df['ElectricityRatio10'] = (df['ElectricityRatio'] * 100 + 9) // 10
df['NaturalGasRatio10'] = (df['NaturalGasRatio'] * 100 + 9) // 10
In [43]:
sns.boxplot(data=df, x='SteamRatio10', y='GHGEmissionsIntensity')
plt.ylim([-1,15])
plt.xlim([0.5,8]);
In [44]:
sns.boxplot(data=df, x='ElectricityRatio10', y='GHGEmissionsIntensity')
plt.ylim([-1,15])
plt.xlim([0.5,10.5]);
In [45]:
sns.boxplot(data=df, x='NaturalGasRatio10', y='GHGEmissionsIntensity')
plt.ylim([-1,15])
plt.xlim([0.5, 10.5]);

We can see that electicity is the cleanest energy source. GHG emissions increase with natural gas and steam use but decrease when electricity is the main source of energy

3.3 Buildings Surface And Geometry

In [46]:
fig, ax = plt.subplots(1,2,figsize=(12,4))

sns.scatterplot(data=df, x='SiteEUI(kBtu/sf)', y='NumberofFloors', ax=ax[0])
sns.scatterplot(data=df, x='PropertyGFABuilding(s)', y='SiteEUI(kBtu/sf)', hue='NumberofBuildings', ax=ax[1]);

Larger building seems to require less energy

A building with multiple floors have less exposed surface area % compared to a one floor building. Threfore, it requires less energy to heat in winter

In [47]:
df['NumberofBuildings'].value_counts()[:2]
Out[47]:
1.0    979
2.0      3
Name: NumberofBuildings, dtype: int64
In [48]:
df['NumberofFloors'].value_counts()[10:11]
Out[48]:
0    10
Name: NumberofFloors, dtype: int64

There is ten building with 0 floors and many without building despite having an energy consumption.

We are going to replace these value by 1 to allow us to create the surface per floor and per building columns

In [49]:
df['NumberofBuildings'] = df['NumberofBuildings'].replace(0,1)
df['NumberofFloors'] = df['NumberofFloors'].replace(0,1)
In [50]:
df['SurfacePerBuilding'] = df['PropertyGFABuilding(s)'] / df['NumberofBuildings']
df['SurfacePerFloor'] = df['PropertyGFABuilding(s)'] / df['NumberofFloors']
In [51]:
df['ParkingRatio'] = df['PropertyGFAParking'] / df['PropertyGFATotal']
df['BuildingRatio'] = df['PropertyGFABuilding(s)'] / df['PropertyGFATotal']

3.4 Correlations between the variables

In [52]:
corr = df.corr()[1:]
plt.figure(figsize=(12,12))
sns.heatmap(corr, annot=True, cbar=None, fmt='.2f', cmap='winter_r');

3.5 Construction Decade

In [53]:
plt.figure(figsize=(12,6), facecolor="#eaeaf2")
df['Decade'] = df['YearBuilt'] // 10
df['Decade'] = df['Decade'] * 10
sns.barplot(data=df, x='Decade', y='ENERGYSTARScore', color='#7490c0');

3.6 EnergyStar Score

One of our goals is to decide if the ENERGYSTAR score is useful.

First we are going to look at its distribution

In [54]:
plt.figure(figsize=(12,6))
sns.histplot(data=df, x='ENERGYSTARScore');
In [55]:
fig = px.scatter(df, x='ENERGYSTARScore', y='SiteEUI(kBtu/sf)', trendline='ols')
pyo.iplot(fig, 'scatter-energystar-site-eui')
In [56]:
fig = px.scatter(df, x='ENERGYSTARScore', y='GHGEmissionsIntensity', trendline='ols')
pyo.iplot(fig, 'scatter-plot-energystar-ghg')

It is difficult to see the correlation between the variables using a scatterplot. We are going to look at the data again using barplots and boxplots to confirm the trend showed by the regression lines

In [57]:
custom_palette = ["#ef3e22", "#f58220", '#feca0a', '#7fc241', "#008b4c"]
sns.palplot(custom_palette)
In [58]:
df['ESScoreGroups'] = (df['ENERGYSTARScore']+19) // 20 #creation de cinq groupes allant de 1 a 5
#les données sont décalées pour éviter la création d'un groupe 0
In [59]:
plt.figure(figsize=(12,6))
sns.barplot(data=df, x='ESScoreGroups', y='SiteEUI(kBtu/sf)', palette=custom_palette);
In [60]:
plt.figure(figsize=(12,6))
sns.boxplot(data=df, x='ESScoreGroups', y='SiteEUI(kBtu/sf)', palette=custom_palette)
plt.ylim([-2,400]);
In [61]:
plt.figure(figsize=(12,6))
sns.barplot(data=df, x='ESScoreGroups', y='GHGEmissionsIntensity', palette=custom_palette);
In [62]:
plt.figure(figsize=(12,6))
sns.boxplot(data=df, x='ESScoreGroups', y='GHGEmissionsIntensity', palette=custom_palette)
plt.ylim([-2, 10]);

We can easily see the correlation between the ENERGYSTAR score and the energy consumption

In the barplot, the average energie consumption is half for the group 5 building, those with the highest ENERGYSTAR scores, compared to the group with the lowest scores. The GHG emissions are 3 times higher for the low ENERGYSTAR scores buildings

4.Final Columns Selection

The last step is saving the data we cleaned and selected. We need to remember to suppress the other energy columns and the ones we used to improve the understanding of data

In [63]:
final = df.drop([
        'SourceEUI(kBtu/sf)',
        'SteamUse(kBtu)', 'Electricity(kBtu)', 'NaturalGas(kBtu)',
        'SteamRatio10', 'ElectricityRatio10', 'NaturalGasRatio10',
        'Decade', 'ESScoreGroups',
        ],
       axis=1)
In [64]:
final.to_csv('seattle_clean2.csv', index=None)